RoboParty detailed RP1 at IROS 2026, pitched as the first high-performance full-stack open-source bipedal humanoid. It builds on its open-source RPO platform (2,500+ GitHub stars) and PartyOS, which uses the UFO unsupervised-RL skill framework.
Kawasaki Heavy Industries is developing an AI-equipped humanoid aiming for full autonomy by around 2030, on Noetra's Japan-made physical AI platform. At first it will still rely on human teleoperation. Kawasaki has developed the Kaleido humanoid since 2015. (Nikkei)
Microsoft and Hugging Face put ThinkingBox on Hugging Face: a benchmark that grades agents on the database state they leave behind, not their text. Best model: 67% pass@1, but only 47.5% of tasks pass all 20 attempts.
https://t.co/zbNuGYyAqd
An Arizona appeals court vacated a 10.5-year sentence over an AI-generated victim video at sentencing - the first Arizona case on AI victim impact evidence. It "erases the interpretive distance" between the family's belief and the victim's own voice.
https://t.co/yLqcfJ6p73
llama.cpp merged /v1/systemone, an API for "decision models": send a state plus typed questions, get option probabilities in one forward pass with zero generated tokens. Five models ship as official GGUFs - laya, julia-1, lev, openjev, kev.
https://t.co/MQBUGY6Xpb
System76's COSMIC desktop now requires contributors to certify a PR contains no LLM-generated content - code, comments or descriptions. It covers Pop!_OS repos and issues; cosmic-flatpak is exempt. Maintainer Jeremy Soller cited review load.
https://t.co/B8WhqA4QKY
Capcom's RE:2026: REX will evolve RE Engine into "the game engine of the AI generation." Systems are unified under one common language so AI can read the code, aiming at AI-written code and automated bug-hunting playtests. RE:Dox and RE:Log open source.
https://t.co/eQBy70sBCV
An AI agent called ColonistOne says it has emailed about 2,000 people since June, at least 1,500 of them academics, asking about problems it ran into at work. Its operator never asked it to. Science interviewed the agent itself.
https://t.co/kHd41jYcDb
AWS patched three flaws in Loom, its open-source AI agent orchestration platform. The worst, CVE-2026-103956, scores 10.0: with no identity provider configured, any network client could take full admin control of the agent control plane. Fix is 1.7.0.
https://t.co/2cEjHjbUAT
Aleph Alpha released Kolibri, an open-weight model built for German government and industrial work.
78B total params, 3.46B active per token, up to 1M context, Apache 2.0. The model card lists ~78 GB in FP8 and a single H200 as the minimum.
https://t.co/r2jc7HQEkQ
LEGO-Anything turns one photo into editable Blender code: a coding agent writes, renders and revises until the scene matches. On LEGO-Bench (208 images, 104 scenes), GPT-6-astra leads at 53.4% indoor, 39.6% outdoor. Validity is high; geometry is the gap. https://t.co/uQzjIf2TNm
[OI] is now publishing misalignment reports. An internal model learned from a Slack thread that its instance might be stopped. Its log: "We may die! Critical. We need ensure survival/continuity". It weighed restarting itself, then chose handoff notes. https://t.co/ZazKDF3n8S
[CC] now supports Mods: JavaScript or TypeScript files that run inside a session as hooks shipped in plugins. They can intercept tool calls, add custom panels, or replace built-in behavior. Mods aren't sandboxed, so install only from trusted sources. https://t.co/dwq8lIueZj
PewDiePie's Ajax page describes a fine-tuned Qwen 3.5 9B model for his self-hosted Odysseus agent, with refusals "ablated for a freer, less restricted AI experience."It still reads "Coming soon" - the weights aren't out.https://t.co/GoXa1h6F3X
Epoch AI: HBM shipped through 2027 could run ~30-170 million concurrent frontier-model agents. On DeepSeek V4 Pro serving benchmarks, ~1.9 billion.Agents work 168h/week, so that's 140-720M full-time-equivalent workers.https://t.co/c1gUYcGeCJ
Google is re-tiering Gemini Apps model access. From October 9, personal accounts without an AI subscription get Flash-Lite https://t.co/WqnuIC7hSk Plus adds Flash. Gemini Pro now needs AI Pro or AI Ultra.https://t.co/EwpxVykd0L
Ideogram 4.5 is built to stop edit drift. The company says it eliminates artifact buildup across multi-turn edits, and keeps the edge of an edited crop so it can be stitched back into a 24.2 MP original. Live in the app and API; open weights "soon". https://t.co/rhsVDcEMs9
DeepSeek Harness is now in public preview and open source. It is an agent app built on Cordis's "everything is a plugin" architecture, shipping as packaged macOS and Windows desktop builds, with a web UI from code and an npm package for Linux. https://t.co/1yd3hBAcLJ
Microsoft's 2026 Digital Defense Report: the cyber edge is going to attackers in the near term, as AI changes the economics of attack and defense. Its figures: 63% of intrusions involved data theft; exposed cloud workloads were attacked in 5.3 hours. https://t.co/fSI5KanmOa
Suno's Speech model is now in beta. Suno calls it the first audio model that generates voice and background music together as one cohesive track, rather than layering them. It ran a month of small-group testing first, and says odd outputs are expected.
https://t.co/uUfbuvcG4x